Two AWS Middle East availability zones down after datacentre fires

Two out of three UAE availability zones experiencing power outages after being ‘impacted by objects’

Two of Amazon’s three AWS Middle East cloud availability zones remain disrupted after a fire at a datacentre in the United Arab Emirates (UAE) on Sunday.

Availability zones are clusters of datacentres with independent power supplies and connectivity located in the same geographical region. They are intended to provide resilience in case of a failure of a single zone.

On its health status dashboard, Amazon says a facility in one zone was “impacted by objects that struck the datacentre, creating sparks and fire” at around 4:30 am PST (12.30 pm GMT) on Sunday. Later it said that two availability zones had been affected by power outages.

In its most recent update (2nd March 02:53 AM PST), Amazon says two of its availability zones in the ME-CENTRAL-1 region are impaired, with key services including Amazon DynamoDB and Amazon S3 “experiencing significant error rates and elevated latencies.”

It adds: “With two availability zones significantly impacted, customers are seeing high failure rates for data ingest and egress.”

The company continues: “We strongly advise customers to update their applications to ingest S3 data to an alternate AWS Region. As soon as practically possible, we will begin the restoration of our two availability zones which will include a careful assessment of data health and any repair of storage if necessary.”

Nine services, including S3, EC2, AWS Lambda and AWS Management Console and CLI have been “disrupted” by the failure of the two availability zones.

In addition, 24 services are currently “degraded”, including DynamoDB, NAT Gateway and Elastic Load balancing, while a further 58 services are described as “impacted”.

Customers are experiencing difficulties launching new EC2 instances, associating Elastic IPs, and managing resources in the impacted zones. Increased error rates for EC2 APIs are reported, particularly networking-related APIs. Meanwhile, EBS volumes and RDS DB instances may be unavailable in the affected zones.

Amazon says it working to route requests away from the affected availability zones, recommending customers use alternative zones or AWS regions while it seeks to restore power.

Amazon expects to provide a new update by 6:00 am PST 2nd March (2.00 pm GMT). Recovery is expected to take several hours.

Amazon does not mention the underlying cause of the fire, but it coincides with an attack on Iran by the US and Israel and retaliatory missile and drone barrages by Iran against their allies, including UAE, Qatar, Bahrain, Lebanon, Kuwait, Iraq and Jordan.